AI for boiler, industrial heating and energy operators.
Built for how the work actually runs: service and maintenance, estimation, field work, energy optimisation, and engineering, procurement and construction (EPC) project delivery. The hard part here is never the model. It is that the knowledge an estimator uses to price a job correctly on your estate (every site and piece of plant you install, service or run) is spread across a field service system, an accounting package, an inbox and one engineer who has been here eleven years.
Fixed-scope phases, run in your own accounts. You own what we build.
We work with any business. The sector pages are worked examples of the method, not limits.
- Field serviceJobs, callouts, technician notes. Usually the closest thing to the truth about what was actually done.
- Asset registerWhat is installed where. Often a spreadsheet, often a version behind the site.
- AccountingWhat it was charged at. Rarely agrees with the quote.
- BMS / telemetryRuntime and combustion data, where it is instrumented at all.
- InboxThe variations, the approvals and the reasons. None of it structured.
- An engineerEleven years of which sites behave badly in winter. Not recorded anywhere.
The marked row is usually the system of record. Usually — that is one of the first things a review establishes, because half the time it is not.
Where the day actually goes.
Scheduling field and project work by hand. Crews carrying paperwork that slows the job and produces patchy data. Delivery risk on projects that is hard to see until it is late. Estimation and admin soaking up the hours of the people who are hardest to replace.
And, almost always, staff quietly pasting customer detail into consumer AI tools because it is faster than the sanctioned path. None of this is unusual. It is what happens as a heating or energy business grows past the point where one person can hold it all.
The work grows with the business. So does the manual load.
What an AI would have to know before it is any use on your estate.
This is the list we work through in a review. Every item is context that exists in your business today, and that a general-purpose assistant has no access to. How estimation precedent turns into a faster quote without losing the estimator's judgement is the subject of Quoting faster without cutting corners.
- 01Plant and asset historyWhat is installed where, what has been serviced, what failed last winter and what was replaced instead of repaired. Usually split between a field service system, a spreadsheet and a senior engineer.
- 02Service precedentWhat this customer has been charged for a similar callout before, what was discounted and why, and which sites have terms that are not in the standard contract.
- 03Estimation precedentThe last twelve jobs of this shape, what they were priced at, what they actually cost, and which assumptions turned out wrong.
- 04Commercial termsResponse times you are committed to, penalties attached, and which accounts you are already carrying at a thin margin.
- 05Current stateWho is on the road, what is in stock, what is overdue, and which project is quietly slipping.
A first automation in 2–4 weeks. A connected estate in 3–5 months.
Assist and automate
A governed assistant (an AI assistant that sees only what each person is allowed to see, and records what it answered) and the first automations on service, estimation and field paperwork. Time handed back while the deeper work is still being scoped.
Connect the estate
The context layer over field service, project operations, finance, email and phone. Fingertip reporting, live dashboards, and alerts on delivery and cash.
Predict and act
Forecasting on demand, capacity, cash and maintenance, plus agents for scheduling, estimation, quoting and field reports. Each starts at L2 Propose; money, safety and people stay at L4.
Autonomy tiers: L1 Assist · L2 Propose (the default) · L3 Act within written bounds · L4 Approval required always. What each tier means
Efficiency signals out of data you already hold.
Once the estate is joined up, the same layer surfaces the signals that drive boiler efficiency: combustion and excess air trends, condensing operation, flue and standby losses, sequencing across boilers, and benchmarking between sites. First for your own plant and reporting, then as a service you can sell back to your customers. Where the real efficiency number lives, and why it needs a connected foundation, is set out in AI for boiler and heating operators.
Thresholds are set with your engineers and validated against real plant, not taken from a textbook. Deeper real-time efficiency work needs sensor telemetry in place, which is a later and separately scoped step. We will say so rather than imply it comes free.
The data was always there. Joining it up is what makes the waste visible.
What we will stand behind.
No projected numbers and no industry averages dressed up as your result. We baseline first, then show the change against your own figures.
Start with a conversation about your estate.
Tell us where the hours go and which systems hold the truth. We will tell you what a first stage would realistically look like, or that there is not one worth doing yet.